Restaurant Demand Forecasting: How to Forecast Sales, Covers and Prep

To forecast demand in a restaurant, start with your POS transaction history: average sales for each day of the week over the last four to eight weeks, then adjust that baseline for seasonality, weather, local events, and reservations on the books. Convert the resulting sales forecast into covers, menu items, and ingredient quantities so it drives prep lists, purchase orders, and staff schedules. Most independent restaurants can reach 80-90% forecast accuracy with a simple spreadsheet and a weekly 30-minute routine — no data science required.

What Is Restaurant Demand Forecasting?

Restaurant demand forecasting is predicting how much business you will do in a future period — revenue, covers, orders per daypart, and sales of individual menu items — from historical data and known upcoming factors. It underpins almost every operational decision: how many cooks to schedule Tuesday night, how much chicken to order for the weekend, how much cash you will have at month end.

The cost of getting it wrong is concrete. Over-forecast and you over-order, over-prep, and over-staff: cash sits in the walk-in, food spoils, labor cost climbs. Under-forecast and you 86 popular dishes mid-service and burn out a skeleton crew. With food cost at 30-35% and labor at 25-35% of sales, there is little margin for either error — which is why forecasting deserves a standing weekly routine, not a gut-feel guess made while writing the order.

What Data to Use for Restaurant Forecasting

Good forecasts combine a baseline from your own sales history with the few external factors that reliably move demand.

POS and Transaction History

Your POS is the single most valuable forecasting input. Export at least 8-12 weeks of sales with item-level detail: what sold, in what quantity, on which day, at what time. Timestamps are especially useful for predicting peak hours — bucket transactions into 30- or 60-minute intervals and patterns emerge fast: the 12:15-13:15 lunch rush, the Friday 19:00-21:00 dinner peak. For more on working with POS exports, see how to analyze restaurant sales data.

Day-of-Week and Seasonal Patterns

A Monday and a Saturday are effectively different businesses, so always forecast each day of the week separately — never from a flat daily average. On top of the weekly cycle sits seasonality: tourist-area cafes can run summer volume 50-100% above winter, while office-district lunch spots dip hard in August. Capture it by comparing the same weeks year over year, or with a simple seasonal index (the month's average sales divided by the year's average).

Weather

Weather moves short-term demand more than most operators expect. Rain can cut terrace-heavy revenue 20-40% on the day; the first warm spring weekend can double it. Checking the 7-day forecast when you build the weekly plan captures most of the value — and logging what actually happened on rainy versus sunny days grounds your adjustments in your own numbers.

Local Events

Festivals, concerts, sports fixtures, and conferences create spikes that history alone will never predict. Keep a calendar of recurring local events and record the sales uplift each produced — that becomes your event multiplier for next year. Road closures, competitor openings, and school schedules belong on the same calendar, since they move demand the other way.

Reservations and Pre-Orders

Reservations are confirmed future demand. Use covers on the books as a floor, subtract your typical no-show rate (often 5-15%), and add a walk-in allowance from history — perhaps 20-30% on a normal night, less when the book is full. Watch booking pace too: if Saturday is filling faster than usual by Wednesday, revise upward before placing orders. Pre-orders and large parties are even better — they give exact item counts.

Restaurant Forecasting Methods, From Simple to Advanced

There is no single correct restaurant forecasting model. Start simple, measure accuracy, and add complexity only when the simpler method's errors cost you money.

Moving Averages

The simplest workable method: forecast next Tuesday as the average of the last four Tuesdays. A 4-week moving average smooths one-off spikes and is accurate enough for stable neighborhood businesses. If sales are trending, use a weighted average so recent weeks count more: (Week1×1 + Week2×2 + Week3×3 + Week4×4) ÷ 10, with Week4 most recent — it reacts faster to real trend changes while still damping noise.

Same Day Last Year, With a Growth Factor

With a full year of history, same-day-last-year forecasting handles seasonality automatically: last year's second Saturday of December already contains the holiday effect. Multiply by your year-over-year growth rate — if you're running +6%, forecast = last year's same day × 1.06. Align by day of week, not calendar date, and adjust manually for floating holidays like Easter.

Per-Daypart Forecasting

A single daily number hides the shape of the day. Splitting the forecast into dayparts — breakfast, lunch, afternoon, dinner — makes it directly usable for scheduling and prep: a €2,400 day that is €900 at lunch and €1,200 at dinner implies a very different staffing plan than an even spread. Build daypart forecasts the same way as daily ones, per time bucket from POS data — the daypart profile is the peak-hour prediction. Our guide to managing peak hours in cafes covers what to do once you can see the peaks coming.

Item-Level and Menu Forecasting

Item-level forecasting — menu forecasting — converts a covers forecast into dish quantities using your menu mix: each item's share of total items sold. If burgers are 18% of items and you forecast 500 items this week, plan for ~90 burgers, then multiply by recipe quantities (90 × 180g of beef = 16.2kg) to get ingredient needs and prep pars. Menu mix is more stable than total volume, so item percentages usually hold even when covers swing. For new dishes with no history, borrow ~80% of the mix of the item they replaced, or start at 10% of a comparable item and adjust after two or three weeks.

How to Build a Weekly Forecast in a Spreadsheet

Here is a process you can set up in under an hour and run in about 30 minutes a week, using nothing but a POS export and a spreadsheet.

Using Forecasts for Scheduling and Purchasing

A forecast pays for itself when it changes decisions — and the two biggest levers, labor and ordering, consume 55-65% of revenue.

For scheduling, work backwards from a productivity target: at €60 of sales per labor hour, a €2,405 Friday means roughly 40 labor hours, distributed to match the daypart curve. Our guide on shift scheduling in cafes covers building the rota, and the labor cost calculator shows what a planned schedule costs as a percentage of forecast sales before you publish it.

Cash Flow Forecasting for a Restaurant

The same sales forecast that drives prep and scheduling also drives your cash flow forecast. A simple 4- to 13-week version takes weekly forecast sales as cash in, then lays out cash out by week: supplier payments on their actual terms, payroll on pay dates, rent, VAT or sales tax, and loan payments.

This matters because restaurant cash crunches are usually timing problems, not profitability problems: a strong month's tax bill, quarterly rent, and payroll can all land in the same week a seasonal dip starts. A forecast showing August running 25% below July surfaces that squeeze six weeks out — time to slow purchasing, trim the schedule, or arrange supplier terms. For the full template, see how to manage restaurant cash flow.

Choosing Demand Forecasting Software

A spreadsheet is the right starting point, but it takes manager hours weekly and can't learn from its own errors. Demand forecasting software for restaurants — built into your POS, inventory platform, or a dedicated tool — ingests transactions continuously, forecasts per item and daypart, factors in weather and holidays, and flags when actuals diverge from forecast.

Software starts paying for itself when manual forecasting eats several manager-hours a week, you run multiple locations, or waste and stockouts persist despite a disciplined process. Below that threshold, a well-kept spreadsheet is genuinely competitive.

Common Restaurant Forecasting Mistakes

• •Forecasting from a flat daily average instead of per day of the week — the most common error; it guarantees over-prepping Mondays and under-prepping Saturdays
• •Using too little history: one or two weeks is noise; use at least four, ideally with the same period last year for seasonal context
• •Ignoring the calendar — a missed festival or holiday causes the worst stockouts and over-orders of the year
• •Overreacting to outliers: one anomalous week is not a trend; note the cause and exclude it from averages
• •Never measuring accuracy, which lets systematic bias (always 10% over or under) persist for months
• •Forecasting revenue only, so the number never reaches the kitchen as prep quantities or the rota as labor hours
• •Copying last week's order regardless of the forecast — if ordering doesn't change when the forecast changes, the forecast is decoration